(AIST Group / Free Webinar) Learning from Research Results of ABCI 3.0 'Development Acceleration Use' ~Building Japanese Speech Foundation Models and Their Application to Speech AI~ [May 25, 28 Online Distribution]

Speech AI technologies, such as speech recognition and speaker emotion analysis, are increasingly being utilized in various situations, including meeting transcription and smart speakers. On the other hand, speech characteristics vary greatly depending on the speaker, emotion, and acoustic environment. Therefore, it has been a major challenge to improve the accuracy of speech recognition and emotion recognition, especially for conversational speech rich in emotional expression and speech from a wide range of generations, due to the difficulty of securing sufficient training data. Speech foundation models have recently attracted attention as a solution to these challenges. By utilizing features obtained from speech foundation models, high performance can be achieved even in environments with limited training data. In this webinar, we will introduce the research and development results utilizing "ABCI," an open computing infrastructure jointly operated by the National Institute of Advanced Industrial Science and Technology (hereinafter referred to as AIST) and AIST Solutions, which supports the bridging of AI technology development and social implementation. The Japanese speech foundation models "Izana*" and "Kushinada," released by AIST in March 2025, are general-purpose models developed using ABCI's computing resources, intended for applications such as speech recognition and speech emotion recognition. On the day of the webinar, we will introduce an overview of the Japanese speech foundation models that AIST is building, how ABCI supports research and development, and specific application examples for speech recognition and speech emotion recognition. This content is recommended not only for those involved in research and development of speech AI and generative AI, but also for those who want to know how to utilize GPU cloud services for their company's AI development, and those interested in the use of computing infrastructure and research and development examples in the generative AI and speech AI fields. *The development of "Izana" is supported by the NEDO (New Energy and Industrial Technology Development Organization) commissioned project "Technology Development Project for Next-Generation Artificial Intelligence that Evolves with Humans (JPNP20006). Recommended for: ・Researchers and technical planning managers who want to grasp the latest trends in speech AI technology, such as speech recognition and speech emotion recognition. ・Researchers and engineers working on the development of Japanese speech models utilizing large-scale speech data. ・Technical planning managers who want to learn practical examples of large-scale model training and distributed learning using GPU cloud services. ・Technical planning managers who face challenges caused by on-site speech data, such as speaker differences, emotions, and noise. ・Corporate managers and planning managers considering the utilization of next-generation AI such as generative AI, foundation models, and multimodal AI. Click here for registration and details: Event Overview Date and Time: May 20, 2026 (Wednesday) 11:00 AM - 11:40 AM May 22, 2026 (Friday) 3:00 PM - 3:40 PM *The content broadcast on both days will be the same. *The end time may vary slightly. Participation Fee: Free Viewing Method: Online distribution. Viewing is possible from a browser. TOPIC 1: Building and Utilizing Japanese Speech Foundation Models with ABCI TOPIC 2: Introduction to the 2026 Support Program Overview AIST Solutions Event Management Team E-mail: webmktg-eve-ml@aist-solutions.co.jp AIST Solutions plans to hold webinars on various themes in the future. We sincerely look forward to your participation. EVENTS/WEBINARS https://www.aist-solutions.co.jp/events_webinars/ Keywords: